Cognitive functioning is a complex theoretical concept, and there is a lack of consensus concerning the definition and use of the term in the literature (7, 34, 35). One consequence of this is the variety of interpretations and conclusions regarding cognitive functioning in depression.

This complexity is sometimes hard to grasp for clinicians and patients outside the field of clinical neuropsychology. As shown in Figure 2, cognitive functioning is a concept consisting of several interrelated sub-concepts defined as domains. Furthermore, within each domain there are several specific aspects. In a traditional neuropsychological assessment, each specific aspect is measured by standardized tests or experimental paradigms. These findings will normally be interpreted and explained on an aspect level in a clinical setting (see Figure 2); however, the literature traditionally describes findings on a domain level, with the risk that important findings on an aspect level will be ignored. When studies report results as composite scores, summarized by scores of different aspects within a specific domain (36, 37), it may lead to a wrong conclusion regarding lack of differences between groups (type 2 error) with the risk that specific impairments will disappear in the composite score. This is of particular importance in clinical neuropsychological work in which the whole ideographical profile is of importance for the individual patient. Consequently, an incorrect portrayal of patients' cognitive profile could be reported, failing to mirror actual challenges in daily life functioning. In addition, using population norms to explain test results (38) may ignore the important ideographic interpretation of how the different cognitive tests are related to the individual patient. In addition, some norms are not standardized for regional conditions and could thus underestimate deficits (39). In addition, a recent meta study by Parkinson et al. (40) argued that results in one domain may not reflect the effects on one test, thus precluding results from single tests.

Composite scores, however, are important in research to identify latent variables representing different cognitive domains, avoiding task impurity problems, and useful for structural equation models investigating cognitive functions above and beyond the clinical setting (7).

With this in mind, the literature from the past decade will be summarized with a focus on cognitive functioning over time, in a long-term perspective, investigating the three previously mentioned hypotheses in particular, in the cognitive domains of memory, executive functioning, attention, and psychomotor processing speed. The concept long-term perspective is used to describe the cognitive profile over time in non-symptomatic phases, which can be reflected in both cross- sectional and longitudinal studies.

Patients with a history of depression often report that they experience memory problems, both in an acute phase (41) and in a recovery phase (42). Based on clinical practice, patients not only describe this problem evidenced in themselves, but also in family members and others. This could lead to increased stress, frustration, relational problems, and negative self-representation (43). Many former patients relate these memory problems to possible brain damage or dementia, leading to a negative impact on self-representation, rumination and coping. Such an interpretation combined with the lack of correct knowledge regarding the role and origin of residual cognitive impairment might lead to an increased risk of relapse and new episodes.

Numerous studies have investigated aspects of memory during the past decade and findings so far appear divergent and non-conclusive. Lee et al. (44) concluded in their meta-study that memory represented a state marker being associated with clinical state in first episode patients. This was also evident in the meta-study by Ahern and Semkovska (45), concluding that first-episode patients showed normalized learning, memory, and autobiographic memory in remission. Pu et al. (46) found deficits in verbal memory but only for a subgroup of patients with MDD. This group also showed deficits in information processing that could influence memory consolidation. This was also the case in Lee et al. (44), where patients with mild depressive symptoms showed second largest impairments in verbal memory, only superseded by processing speed. An older MDD group showed deficits in visual and verbal, learning and memory, but also in most cognitive functions. Xu et al. (47) reported immediate visual memory impairment (WMS-R Immediate Visual Reproduction; copying figure after 10 s exposure) in patients in the depressed state and in remission compared to healthy controls, suggesting visual memory deficits may be a trait for mood disorders. This finding, however, was not supported in a study conducted by Hammar and Schmid (48), whose findings indicate that visual memory performance in patients with major depression normalizes during a 9-month period, thus indicating a state-related relationship. This study had a small sample, suggested a deficit in copying at follow up, and lacked the immediate condition that could be most analogous to Xu et al. (47). Thus, a trait/scar deficit in visual construction is indicated. Hammar et al. (49) suggested that MDD patients showed intact verbal memory performance, a deficit in immediate verbal learning, and deficits in visual memory in the symptomatic phase. Immediate verbal learning deficits were still evident during partial remission (50). Shimizu et al. (51) reported that remitted MDD patients had poorer verbal memory, both immediate and delayed (as well as deficits in most other cognitive areas), compared to healthy controls. This was in line with findings showing that despite significant improvement in memory from the symptomatic phase to remission, patients were still significantly impaired compared to healthy controls (52). Wekking et al. (53) found impairment in several measures of memory in remitted patients but could not establish that the impairment predicted future relapse within in a 24-month period. Vasavada et al. (54) found deficits in verbal learning before ECT treatment, but also following symptom reduction, suggestive of trait or scar effects. A study of remitted patients and their high-risk twins (55) did not find any verbal memory impairments in either group, contrary to what would be expected from a trait perspective. However, a recent meta-study found small impairments in most cognitive functions in first-degree relatives of patients with MDD, including in memory, supporting this hypothesis (56). Finally, studies investigating first-episode patients and patients with recurrent depression with biological correlates have supported a scar hypothesis: Hansson et al. (25) reported no relationship between an abnormal HPA axis and cognitive dysfunction in verbal memory in first-episode patients. However, abnormal HPA-axis and cognitive impairment was evident in patients with a history of recurrent depression, indicating that the episodes might have a scarring effect on verbal memory functioning (24). Tully (57) also finds potential scarring effects, with depressive symptoms and high blood pressure associated with decline in visual memory. This view is also supported by a review by Mcintyre et al. (58), that posits that a subgroup of individuals with MDD show progressive decline in memory. A study on late-life depression suggested that depressive symptoms were a prodrome for Alzheimer (59), that could suggest a dementia state effect on reduced memory in late life MDD (see Table 1).

In conclusion, the literature supports evidence of a long-term memory impairment in depression. This is not independent of attentional and learning deficits, as evidenced by the sustained difficulties with immediate memory, and it could therefore be influenced by impaired informational encoding more than a long-term memory deficit. In addition, all three hypotheses concerning role and origin have partial support. Hence, the neurocognitive memory profile in depression is neither specific nor conclusive and requires a multidimensional approach. This domain is of particular interest with regard to development of neurodegenerative disorders and is often affected first in the development of Alzheimer's disease. Studies seem to find increased deficits in memory related to depression with increasing age (14, 60).

Patients who have experienced or are experiencing depression frequently report that they often have difficulty performing tasks that require initiating or finishing activities, problem solving or getting an overview of situations, multitasking, or emotional regulation; inhibiting negative or troubling thoughts (41–43). We clinically relate these tasks to a higher order of functioning for regulation of behavior, thoughts and emotions –defined as Executive Functions.

Aspects of EF have been investigated in several studies, and findings tend to highlight inhibition, defined as suppressing an automatic response in order to make a less automatic but task-relevant response (61). The concept of inhibition has been operationalized differently in different studies depending on the neuropsychological task used (7). Lee et al. (44) suggested in his review that inhibition could be a trait marker in first-episode patients. This conclusion was supported in findings from several studies conducted by Schmid and Hammar (62), who stated that impaired inhibition on the stroop test, in addition to semantic fluency, is present early in the course of MDD, indicating that EF represents a trait in MDD, irrespective of symptom severity and number of previous episodes. Moreover, the authors showed that impairment in inhibition and switching and semantic fluency in first-episode MDD persisted in long-term follow up (63), with the former associated with relapse during the first year after the first episode (30), and with deficits in inhibition in a subgroup with relapse 5 years later (64), suggesting a relationship between impaired ability in EF of inhibition and switching and relapse in MDD. These findings were also evidenced in patients with recurrent depression and showed that impaired inhibition in the acute phase persisting in phases of symptom reduction (65). Another study on the same patient group showed that impaired inhibition in the symptomatic phase was strongly correlated with impaired inhibition in long-term follow up, indicating that this may represent stable a trait marker in recurrent MDD (66). Moreover, one of the longest follow-ups investigating the same recurrent MDD patients and controls over a 10-year period showed that patients were still impaired with regard to inhibition (13). A twin study did found no EF deficits, neither affected- nor high-risk twins (55). However, a meta study of cognition in first-degree family members did find small effects for deficits in EF (56). This strengthens the hypothesis of at least some trait relation. In addition, findings regarding neural correlates and impaired inhibition were reported in partially remitted and remitted-recurrent MDD patients showing hypoactivation in striatal areas (67). These findings could be interpreted as results of scars or being trait-related. This is in line with findings from Peters et al. (19), who concluded that impaired inhibition as cognitive control in acute and remitted states may represent a trait vulnerability or an early course scar of MDD viable target for secondary prevention or cognitive remediation. Also, Bora et al. (60) concluded in their meta-analysis that response inhibition seems to be a persistent feature in adult-onset MDD, thus supporting the trait hypothesis. They reported that among all cognitive functions, inhibitory control showed the largest magnitude of observed deficits in euthymic MDD patients compared to controls. However, in contrast to all previous findings, Aker et al. (68) reported no deficits in cognitive inhibition in remitted patients, only in response inhibition. Wekking et al. (53) also found persisting deficits in most cognitive measures, except inhibition. Both studies used a contrast score that only approached statistical significance and results similar to other studies (30, 63).

Other studies have focused on different aspects of EF. Contrary to several conclusions regarding trait-related explanations, Roca et al. (69) showed normalization in several cognitive measures such as problem-solving in first-episode and recurrent-episode remitted patients; however, they did not find improved inhibition in the sample. Still, they conclude that remission, rather than numbers of previous episodes, has a high impact on cognitive performance in MDD patients, thereby supporting a state model. Other studies consistently find EF impaired in the depressed state (70, 71), with some inconsistencies with regards to improvements with symptom reduction (54, 72, 73). Pu et al. (46) found small correlations between an EF composite and depressive symptoms; however, this could vary by specific EF tasks measured (63). Age could also influence EF deficits, Boedeker et al. (74) found impairments in switching, but no statistically significant differences in inhibition, in an aging MDD sample. Maalouf et al. (75) using a planning task in adolescents with acute and remitted MDD and they stated that planning and impulsivity appear to be state-specific markers of MDD in adolescents, and are related to depression severity and are not persistent in remission, but a relatively small- and poorly matched sample as well as tasks used could explain this. Early experiences could influence cognition, and Saleh et al. (76) found worse EF in a MDD group with early life stress. Chakrabarty et al. (77) found that only a MDD population with trauma showed persisting deficits in WM in remission. Albert et al. (36) found a relationship between longer duration of depression age, and EF with no effects of current depression severity on performance. The authors concluded that cognitive performance worsens with recurrence over the life span. These findings can be interpreted as support for the scarring hypothesis. Bhardwaj et al. (78) drew the same conclusion, demonstrating an impairment in a planning and problem-solving task in recovered MDD patients and found that performance was correlated with number of previous episodes of depression. They concluded that impairments of EF are present in recovery and are thus not simply state markers, but instead scars caused by previous episodes (see Table 2).

In sum, existing research supports the assumption of a long-term impairment within the EF domain in general, evident in inhibition in particular, with evidence indicating a trait-related profile. However, all three hypotheses regarding role and origin in EF were supported. Similarly to findings within the memory domain, the neurocognitive profile of EF in depression is neither specific nor conclusive.

Depressed patients and formerly depressed patients often report problems maintaining attention during conversations or when reading a book or watching TV, etc. These problems have an impact on daily life functioning and may frequently be interpreted as ignorance rather than the result of a cognitive impairment related to the depression or as a residual cognitive symptom (41, 42). This is clinically related to the domain of attention.

Several studies confirm that attention deficits are related to depression both in the acute phase of the illness (2, 44) and in remission (51, 79, 80). Ji et al. (72) found improvements in digit span in remission, that could indicate a relationship between MDD and attention. In addition, another study found an association between attention and inflammatory markers, which could partly explain state effects in attentional deficits. Findings are divergent, however, some studies report no attentional deficits measured by digit span in MDD in patients both in the acute depressive state and in remission (47). Consistent with this, Boedeker et al. (74) did not find deficits in digit span in an aging MDD sample. This could reflect heterogeneity in the neurocognitive profile. Studies that separate subgroups of MDD in first-episode and recurrent-episode showed that first-episode patients differed in neurocognitive profile: While first-episode patients demonstrated no impairment in attention in effortful information processing in the symptomatic phase of depression (81), the group suffering from recurrent episodes showed an impairment in symptomatic and symptom reduction phases, which normalizes over a 10-year period (82). Such findings might support the Scar hypothesis, showing that duration could be a critical cue to the attentional deficit. These findings were done with a novel paradigm measuring visual attention and might not be generalizable to all other aspects of attention, however. Differences in age between FE and recurrently depressed could perhaps explain this (83). Pu et al. (46) also found subgroups with differing deficits, with one showing among other cognitive deficits, impaired attention. Clery-Melin and Gorwood (84) found differing outcomes supporting the Trait hypothesis; they showed that attention measured by omission mistakes was an unchanged marker before and after treatment and could predict clinical and functional outcome. They interpreted their findings as reflecting a specific ability to control attention and thereby regulate emotional stimuli, thus representing a trait resilience marker, meaning that patients with enhanced ability in attention are more likely to achieve full clinical and functional remission. Attentional control could arguably be considered an EF, however. In addition, this could illustrate the differences in attentional tasks between error scores and RT. The trait perspective was also supported in a meta-analysis conducted by Lee et al. (44), in which they concluded that attention is more likely a trait-marker in first-episode patients. Attention is a complex cognitive domain and is highly interrelated with the other cognitive domains such as memory, EF and psychomotor tempo and could thus influence all other domains (35) (see Table 3).

In sum, an update from the past decade on attentional deficits in depression shows that several aspects of attention are affected, both during the depressive episode and as a residual symptom. The role of attention deficits in relapse and development of new episodes is still unclear and impairments in this domain probably influence results in the other domains (35).

Sometimes patients with previous episodes of depression state that they need more time to complete tasks compared to earlier, this is something we often define as processing speed, psychomotor tempo or information processing. In the clinical setting, it could be labeled latency time and can be quite severe in some severely depressed in-patients (42). Processing speed is consistently impaired in the acute phases of MDD (33; 54) thus supporting a state perspective. Pu et al. (46) however, found only minor relationships between motor-speed and depressive symptoms. Zhang et al. (86) also found slower improvements in processing speed rather than depressive symptoms, suggesting persisting deficits. Albert et al. (36) found a composite measure of processing speed to be the most impaired in MDD. Similarly to EF above, the authors find an interaction between age, duration of depression, reduced processing speed, although current symptoms of depression did not influence this processing speed (when controlling for age, race, sex, educational level and medical comorbidity), thus supporting the trait and scar hypothesis. Meluken et al. (55) did not find deficits in processing speed in relatively young MDD population and related twins, in contrast to state and trait perspectives. The study from Saleh et al. (76) suggest that early traumatic experiences could influence processing speed in MDD, and Chakrabarty et al. (77) found that a MDD population with trauma showed persisting deficits in processing speed following remission. In addition, Jaeger (87), in a review of the digit symbol substitution test, i.e., a measure of processing speed, cites research that finds consistent impairments in processing speed and effect sizes that increase in elderly MDD populations. This is supported in Boedeker et al. (74), who found deficits in an elderly MDD population. Other studies have shown that patients in remission of recurrent depression also suffer from impairment in processing speed (47, 51, 53). Xu et al. (47) did, however, find the most substantial improvements on measures of processing speed. This is in line with Schmid (30, 62). Egerhazi et al. (52) also observed an improvement in psychomotor speed during remission. Vasavada et al. (54) found improvements only in processing speed. Meta studies support this: Patients with first-episode MDD showed an impairment in psychomotor speed in the depressive state and the authors concluded that this deficit was associated with clinical state (44). Ahern and Semkovska (45) also reported in their review and meta-analysis of first-episode depressed patients that remission was associated with normalization of function in processing speed. However, different subgroups could show different impairments (46) (See Table 4).

In sum, processing speed seems to be the most impaired aspect of cognition in depression, but is also most influenced by state trait (and scar) effects. Subgroups in MDD could show more impairment. Results regarding processing speed in MDD deviate in several instances; however, altogether, it seems that recurrent patients show a prevalent slowing in processing speed, whereas first-episode patients show normalization of speed in remission. This pattern might indicate a scaring effect on speed, but also effects of increased aging.

Because of the association between cognitive residual symptoms and the risk of relapse and new episodes, we have to invent treatment programmes (93) targeting these symptoms both acutely (94), and in remission (42, 95). By targeting the cognitive residual symptoms with interventions that remediate or enhance the cognitive capacity one might prevent the negative loop as consequence of failing to function optimally in everyday life (96, 97).

Cognitive enhancement therapy (CET) comprises three important elements; (1) psychoeducation of cognitive residual symptoms, (2) strategies and training of cognitive residual symptoms, (3) transfer the skills to everyday life functioning (96). There are however, several major challenges that must be addressed before such interventions can be standardized treatments of cognitive residual symptoms. First, knowledge regarding cognitive residual symptoms has been acknowledge and understood among healthcare personal and has to be incorporated both in education and in therapist training. Secondly, the frontiers of such interventions have to be explored in open and full trials with specified outcome measures, with a clear goal of enhancing the cognitive capacity in this patient group with a transfer to everyday life functioning. Thirdly, such interventions must be available to the patient group, which is normally outside primary care, since end of treatment for depression is often set at remission of mood symptoms. One way to achieve this is to make CET available in e-health care. Conclusions from a recent open pilot study of an internet-delivered CET intervention showed high compliance and feasibility in such an approach, besides the fact that the remitted MDD patients reported significantly less cognitive residual symptoms after the intervention and that this improvement prevailed at 6 months' follow-up (98). Through in-depth knowledge regarding neurocognitive profiles of depression, it will be possible to target specific aspects in CET treatment to prevent chronic course, disability, and potentially reduce the incidence of dementia. Recent and high-quality evidence on the effectiveness of cognitive-oriented psychosocial interventions has been provided in the treatment of other mental disorders characterized by cognitive impairment, e.g., schizophrenia (99) and one might expect that these promising findings may also be applicable for remitted MDD patients with cognitive residual symptoms.

It is important to note that the present study is not a systematic review. It is based on a comprehensive literature search and is intended to present a narrative review to identify research gaps in the field and highlight methodological concerns. This, however, comes with the risk of not being able to clarify issues such as the future research questions that are not needed (100). Moreover, this summary has not found any cohort studies measuring cognitive functioning prior to first episode of depression, which is the ideal design for support of the trait hypothesis. Following this, the presented literature supporting the trait hypothesis should be interpreted as tentative.